Modeling Technology Transitions under Increasing Returns, Uncertainty, and Heterogeneous Agents

1
Modeling Technology Transitions under Increasing
Returns, Uncertainty, and Heterogeneous Agents

• Tieju Ma
• Transition to New Technology (TNT)
• International Institute for Applied Systems
  Analysis

2
Three missing stylized facts in traditional
technological change models

• Increasing returns to adoption (Endogenous
  technological learning)
• Uncertainty
• Heterogeneous agents following diverse technology
  development and adoption strategies.

3
Technological learning (Increasing return)
Reductions of investment costs for three
representative new and advanced
technologies Source Nebojsa Nakicenovic,
Technological change and diffusion as a learning
process
4
Uncertainty
Range of Future Investment Cost Distributions
from the IIASA Technology Inventory for Biomass,
Nuclear, and Solar Electricity-Generation
Technologies, in US(1990) per kilowatt
(KW). Sources Messener and Strubegger (1991)
Nakicenovic et al. (1998)
5
Heterogeneous agents (actors)

• Traditional model assume a global social
  planner
• In reality, there are different actors with
  heterogeneous attributes, e.g. different attitude
  to risk.

6
Purpose

• Model endogenous technology transitions under the
  three important "stylized facts" governing
  technological change.
• The main objective of the model is for
  exploratory modeling purposes and as a heuristic
  research device to examine in depth the impacts
  of alternative model formulations on the
  endogenous technology transition dynamics.

7
A highly stylized model-- Inspired by energy and
climate change policy models

• One primary resource, whose extraction costs
  increase over time as a function of resource
  depletion.
• One homogeneous good, the demand for which
  increases over time.
• Three technologies
• Existing -- entirely mature, constant cost and
  efficiency, high emission
• Incremental -- slight efficiency advantage,
  higher initial cost (2), potential for
  technological learning (10), low emission
• Revolutionary -- requires no resource input, much
  higher initial cost (40) , higher learning
  potential (30), no emission
• Optimization model

8
Uncertainty in the model

• Uncertain learning rate the learning rates are
  treated as random values characterized by a
  distribution function.
• Uncertain carbon tax. The existence, magnitude
  and the timing of introducing carbon tax are
  treated as uncertain, characterized by different
  distribution functions.
• We generate N sample of random variables, and
  then the average cost resulted from
  overestimating or underestimating the variables
  is added into objective function.
• Solutions are optimal hedging strategies against
  risk.

9
Simulations with one agent
Deterministic learning
10
Historical technology substitution patterns
Competition among multiple technologies. The
share of steel production in the United States by
five different methods. From 1850 to
Source Nakicenovic (1990) Grubler etc (1999)
11
Pareto optimization with two heterogeneous agents
Pareto Optimality The "best that could be
achieved without disadvantaging at least one
group." (Allan Schick, in Louis C. Gawthrop,
l970, p.32)

• Different risk attitude and different weights
• Trading on good
• Trading on resource
• Technology spillover

12
Simulation with two agents and technology
spillover
Pioneer
Follower
13
Diffusion pattern in real world
Diffusion between leading and laggard
markets Source Grubler and Nakicenovic (1991)
14
Carbon abatement
15
Concluding remarks

• The highly stylized model and simulations can
  enhance peoples imagination about how the three
  stylized facts impact technological change
  processes.
• In addition, the simulation results can give some
  policy implications for both risk-taking and
  risk-aversion decision makers, e.g., for
  risk-aversion agent, it is better to import a new
  technology from risk-taking agent at the niche
  market stage of the new technology, instead of
  waiting until the new technology being mature.
• History (story) -based VS Equation-based

16
Thanks for your attention!